3.1—Introduction
Researches are increasingly interested in using FUS for neuromodulation [117]. FUS provides a method to noninvasively target millimeter scale regions of the brain for
neuromodulation. FUS can also influence neural activity by targeted drug delivery [6], [64] or ablation of regions of the brain which contribute to disease [52]. For all of these applications of transcranial FUS it is important to accurately target the brain region of interest and to know where the focus is located during treatment.
For clinical ablation procedures, a small temperature rise is induced in the tissue [3] and its location is measured using MR thermometry [105]. However, too much heating at either the target or in an off target region, which may not be within in the image field of view, can cause damage to otherwise healthy tissue [51], [118], [119]. The brain is also known to be sensitive to temperature changes [120] and for neuromodulation studies the neurological changes induced by the targeting pulse will be mixed with any effects of the neuromodulation pulse. Optical tracking has been used to target brain regions [10], [57]. While optical tracking provides an estimate of the transducer location, it is offline and typically only overlays a free-field estimate of the transducer focus. A real-time method to visualize the acoustic focus that would not cause bioeffects or confounds would be desirable.
1 The work described in this chapter was part of shared first author publication [91]. MAP explored the safety considerations of the FUS pulses using simulations and MR thermometry and ARFI and SVJ designed and
MR-ARFI is a promising method of localizing the FUS beam for neuromodulation. MR-ARFI uses motion encoding gradients (MEG) to image the displacement induced by the FUS beam [25].
These displacements are caused by the acoustic radiation force acting on the tissue. Since the acoustic radiation force is proportional to the intensity of the FUS the displacement maps generated by MR-ARFI provide an accurate map of the FUS location within the brain. MR-ARFI has been shown beyond an ex vivo human skull [121] allowing the beam location to be
measured for transcranial applications of FUS.
An ideal targeting and visualization method for transcranial FUS would not cause any bioeffects at the target or in off target areas. There are two major sources of potential damage with ultrasound: heating and cavitation [122]. As an ultrasound wave moves though tissue some of the energy is absorbed and converted to heat (see section 1.3.2). The amount of heating depends on the ultrasound parameters used and the tissue type. The attenuation coefficient of the skull (up to 20 dB/cm/MHz) is much larger than the attenuation coefficient of brain tissue (approximately 0.6 dB/cm/MHz) [15]. The rarefaction of FUS waves can also cause cavitation in tissues where bubbles are formed and collapse and these bubble collapses can damage tissue. For these reasons it is important to limit the FUS energy as much as possible to minimize potential damage or off target effects of the ARFI targeting pulse.
As part of a larger project studying the effects of FUS on the tactile circuit in a NHP [10], MR-ARFI methods were developed and the safety of the FUS pulses was studied. Simulations were used to estimate the pressure and heating during MR-ARFI sonication pulses and MR thermometry measurements were made during in vivo sonications to assess safety. The in vivo displacement was measured at decreasing FUS pressures to determine the lowest energy
needed for localization. By analyzing thermal deposition and pressure and comparing with observations made with MRI, our study highlights important parameters to consider to minimize bioeffects when mapping ultrasound beams with MR-ARFI.
3.2—Methods
3.2.1—Pressure estimates
A spherical cap single element transducer (Sonic Concepts H115MR) was used for all experiments. The radius of curvature is 63.2 mm and the aperture is 64 mm. The transducers center frequency is 250 kHz and was operated at its third harmonic of 802 kHz. The transducer was placed in a 3D printed coupling cone which could be filled with water and had a 3 cm aperture. Acoustic measurements were made with a needle hydrophone (HNC 0400, Onda Corp., Sunnyvale CA). The free-field pressure from the FUS transducer was measured for a series of input voltages up to a MI of 1.2 and a calibration curve was determined from this data set which was used to estimate pressures used for simulations and MR-ARFI.
An ex vivo skull fragment was used to estimate pressure loss due to the skull. The skull fragment is approximately 7 × 6 × 3 cm in dimension with a thickness between 2 and 3 mm and is from the top of the skull. The fragment was placed in degassed water for 24 hours prior to the measurements to rehydrate the fragment. The FUS transducer was coupled to the side of a water tank through a thin plastic acoustic window, and the needle hydrophone’s voltage was recorded with the transducer driven at 802 kHz. To account for variations in thickness and incident angle the pressure during transcranial sonications was measured at the free-field focal location for five different placements of the skull. The transmission percentage was taken to be
the ratio of pressure measured with and without the skull present. This transmission percentage was used to derate our free field pressure values to estimate the focal pressure within the brain for in vivo studies.
3.2.2—Pressure and thermal simulation
The k-Wave package [93], [94] was used to model the NHP skull fragment surrounded by brain tissue for pressure and thermal simulations. The skull fragment was placed in degassed water for 48 hours to rehydrate and then embedded in 1% agar. A CT scan of the NHP skull fragment was acquired on a clinical PET/CT scanner (Philips Vereos PET/CT, Philips Healthcare, Best, NL). The scans were collected at 140 kVp and 300 mAs and with an in-plane resolution of 0.19×0.19 mm and a slice thickness of 0.67 mm. The data were reconstructed with soft tissue (filter type ‘B’) and bone (filter type ‘YC’) filters and the reconstructions were in HU. For simulations, the image volume was resampled to isotropic 0.3 mm voxels using the imresize3 function in MATLAB (Mathworks, Natick, MA, USA).
The dynamic range of the CT data was compressed so that the resulting speed of sound and density maps better resembled previously reported values. To determine the high range of data A 400 bin histogram (the bin size was approximately 7 HU but varied depending on initial HU minimum and maximum values) was generated. The HU value from the highest bin with at least 500 voxels was set to be the maximum HU for the skull. The CT data was then compressed so that all values below 0 HU were mapped to 0 HU and all values above the maximum
threshold were remapped to that value (1632 HU for the bone reconstruction filter). The soft tissue filter was used to generate a mask of the skull fragment. This mask was used to generate
the parameter maps for acoustic absorption, thermal conductivity, and specific heat in the simulations (table 3.1). The bone filter reconstruction data was used to generate the parameter maps for speed of sound and density using a method similar to Aubry et al. [99]. The porosity was estimated for each voxel as ϕi = 1 − (HUi/max(HUvolume)). This value was then used to calculate a speed of sound and density for each voxel ρi = ϕ*ρwater + (1 − ϕ)*ρbone; ci = (cMax − cMin)*(1 − ϕ) + cMin where ρwater = 1000 kg/m3, ρbone = 2100 kg/m3, cmin = 1500 m/s1, and
cmax = 2900 m/s1. All grids representing the propagating medium were padded to a grid size of [Nx,Ny,Nz] = [300,280,280].
Table 3.1: Tissue properties used in the simulation grid
Parameter Skull Non-skull
Absorption (dB/cm/MHz) 8 0.4
Thermal Conductivity (W/m/K) 0.3 0.5
Specific Heat (J/kg/K) 1700 3600
The H115MR single element transducer was modeled in k-Wave so that the geometric focus was approximately 1cm past the inner surface of the skull fragment. The simulations shown in this chapter were computed retrospectively after acquiring in vivo data. An 80-cycle pulse was used as the input source for the transducer and the amplitude was set so that a free- field simulation resulted in the same pressure that was estimated for in vivo (methods
described below). An additional pressure simulation was performed to compare our simulation results with our water tank measurements behind the skull fragment. For this simulation the
free field pressure was matched to the measured free field pressure in the water tank (284 kPa) and then a simulation was performed through the skull to simulate the transmission loss in a water tank. A GPU-accelerated 3D k-Wave simulation was run on a workstation PC (HP Z820, Xeon E5, with 256 GB RAM, Hewlett Packard, Palo Alto, CA) with a 16 GB Nvidia Titan GPU (Nvidia, Santa Clara, CA). The maximum pressure was recorded for every voxel in the simulation grid.
Thermal simulations were performed using the k-Wave package to solve Pennes’
bioheat equation. The maximum pressure from the pressure simulation and acoustic
absorption were used as the volume rate of heat deposition for each voxel. Due to computation constraints the thermal simulations were only performed in a region around the focus that constrained the part of the skull fragment that experienced the highest pressure. The thermal simulation grid size was [Nx,Ny,Nz] = [140,80,80] with 0.3 mm spacing in all dimensions. The heat source was turned on and off to simulate the FUS pulses used for MR-ARFI with 4.5 ms pulses with a 1 Hz PRF. A time step of 0.1 ms was used during the 4.5 ms ARFI pulses, while a longer time step of 5 ms was used during the “off” periods to ensure feasible computation time. 100 sonications were simulated and the temperature was recorded at each point in the grid for each time point.
3.2.3—MR imaging
MR thermometry was used to measure heating in vivo during MR-ARFI pulses.
Sonications consisted of 4.5 ms pulses repeated at 1 Hz. A gradient recalled echo sequence was used with a FOV of 10 cm by 10 cm, 2 mm isotropic voxels, 5 slices that were 2 mm thick, and
TE/TR 10/25 ms was used. The phase data were converted to temperature change with the follow equation
Δ𝑇 = 𝜙(𝑇H) − 𝜙(𝑇)) 𝛾𝛼𝐵)𝑇𝐸
Eq. 3.1
where 𝜙(𝑇H) is the phase at the current time, 𝜙(𝑇)) is the phase at a baseline before heating was initiated, 𝛾 is the gyromagnetic ratio, 𝛼 is the PRF shift coefficient, 𝐵) is the magnetic field strength , and TE is the echo time [105]. Two regions of interest were selected, one around the focal location and one in the muscle above the skull to estimate skull heating.
To minimize the necessary energy to detect displacement signal the MEGs of the MR- ARFI sequence could be rotated to match the transducer using optical tracking information (for more detail this sequence worked see chapter 4 of [123]). A phantom was constructed to allow for rotation of the FUS transducer during MR-ARFI scans (figure 3.1). The phantom was a graphite-agar combination (4%/1% mass per volume) which displaced during MR-ARFI sonications. This phantom was physically connected to the coupling cone of the transducer which was filled with a 1% agar gel and a 6cm surface coil was positioned at the end of the
Figure 3.1: The phantom experiment allowed the transducer to be rotated relative the magnet coordinate system. The MEGs could be rotated to align with the beam propagation direction to measure the displacement at different MEG orientations.
coupling cone around the phantom. MR-ARFI images were collected with the MEGs aligned in each axis of the magnet coordinate system and in the transducer direction as measured by optical tracking.
MR-ARFI images were collected with a FOV of 12 cm by 12 cm, 2 mm isotropic voxels, a single 2mm thick slice, and TE/TR of 17/1000 ms. The MEG were 3ms in duration with the scanners maximum gradient strength of 40 mT/m and a 4.5 ms FUS pulse was triggered 2 ms before the MEG. Four sets of phase images were collected to reconstruct the ARFI image, 𝜙•Ž• wx‘ , 𝜙•Ž• wOO‘ , 𝜙•Ž• wxM , and 𝜙•Ž• wOOM where the + and – refer to the polarity of the MEG.
Displacement maps could be measured from these phase images by complex phase subtraction Δ𝑥 = 𝜙•Ž• wx‘ ∙ (𝜙•Ž• wxM )∗ ∙ (𝜙•Ž• wOO‘ ∙ ”𝜙•Ž• wOOM •∗)∗)/2𝛾𝐺𝑡 where 𝛾 is the gyromagnetic ratio, G is the gradient strength, and t is the gradient duration.
3.2.4—In vivo imaging
In vivo MR-ARFI was performed on two healthy adult macaque monkeys (M fascicularis) with the approval of the Institutional Animal Care and Use Committee (IACUC) at Vanderbilt University and in accordance with all relevant guidelines and regulations. The NHP was initially sedated with ketamine hydrochloride (10 mg/kg) and atropine sulfate (0.05 mg/kg) and then anesthetized with isoflurane (1.0–1.5%) delivered over oxygen. After intubation, the animal was secured by ear bars, eye bars and a mouthpiece in a custom-designed MR stereotaxic frame. A 2.5% dextrose in saline solution was infused intravenously (3 ml/kg/h) to prevent dehydration during the scans. Animals were artificially ventilated throughout the experiment. Temperature was maintained by a water circulating blanket. Heart rate and peripheral capillary oxygen
saturation (SpO2; Nonin), respiration pattern and end-tidal CO2 (24–32 mmHg; SurgiVet) were continuously monitored and maintained during the entire scan period. The transducer was aimed at the somatosensory cortex, a target which was part of an ongoing neuromodulation experiment [10], [65]. Two experiments were performed to assess reduced energy, a set of rotated MEG gradients (aligned by optical tracking, 45° off, and 90° off) for an in vivo test of the MEG rotational response and a set of three different sonication pressures (free field PNPs of 1.68, 2.25, and 2.81 MPa) were used to test how much energy is needed to detect the focus.
3.3—Results
3.3.1—Transducer output and simulations
The output of the FUS transducer was measured with and without the ex vivo skull fragment in the beam path the characterize the attenuation of the beam due to the skull. The mean detected pressure at 802 kHz in a water tank was 284 kPa (free-field) and 91 kPa (skull
Figure 3.2: Pressure simulations with and ex vivo NHP skull fragment. (A) The simulation grid showing a CT of the NHP skull fragment, the transducer position at the top in yellow, and the three axis view through the focus location where thermal simulations where performed. (B- D) Peak pressure maps around the focus in three views. The highest pressure in the
simulations was inside the skull but the pressure maps are scaled to better show the focus.
fragment present). With the skull fragment, 32.1% of the free-field pressure was transmitted to the same location beyond the skull. Simulations matched to these water tank experiments resulted in a free field simulated pressure of 284 kPa and transcranial pressure of 118.5 kPa.
This simulation showed 41.73% transmission through the ex vivo skull fragment. Simulations of Figure 3.4: Thermal simulation of 100 MR-ARFI sonications. (A) The simulations grid showing a pressure simulations overlaid onto the NHP skull fragment CT. The red ROI shows the approximate area of maximum heating at the focus and the blue area shows the maximum heating area within the skull. (B) The simulated temperature rise at the focus was less then 0.1C . (C) The temperature rise within the skull was approximately 2C. The heating from the skull can also spread to the outer regions of the brain.
Figure 3.3: Temperature rise in vivo during 300s of a MR ARFI FUS pulse. A FUS pulse was applied at 1Hz and the temperature in the brain and in the muscle above the skull where measured with MR thermometry. Similar small temperature rises were seen in both the muscle and the brain.
the same FUS pulses used during subsequent MR-ARFI showed a maximum pressure at the target in the brain of 1.14 MPa and a maximum pressure in the skull of 3.19 MPa (figure 3.2, note the color map is scaled to emphasize the focus in the brain).
Thermal simulations showed that the skull, which absorbs more sound than brain tissue and experienced the highest pressure, had the largest thermal rise with only modest
temperature increases in the brain (figure 3.3). A 100-sonication simulation of an ARFI pulse resulted in a maximum heating in the skull of 2.22 °C and a maximum heating at the target of 0.05 °C. The heating in the skull approached a steady state; however, at the focus, the heating did not appear to be in a steady state.
Figure 3.5: Top: An example of a single rotation of the phantom experiment showing the displacement maps generated with the MEGs aligned in the three magnet coordinates and with the optically tracked beam direction. Bottom: The maximum displacement was detected when the MEGs were aligned with the beam propagation direction.
3.3.2—In vivo imaging
In vivo MR thermometry showed similar, minor heating in the muscle above the skull and the focal area in the brain of approximately 0.3 °C (figure 3.3). These measurements do not represent the true skull heating since there is not enough MR signal from the skull and each ROI contains tissue that is near the skull from which heat can spread into the tissue.
In phantom scans MR-ARFI scans with different MEG angles the maximum signal was detected when the MEG align with the transducer beam propagation direction. Figure 3.5 shows average displacement over a 3x3 voxel region of interest at the focus location for 5 different transducer angulations with different MEG orientations. When measuring along the magnet coordinate system, the signal is maximized when the transducer is aligned with one of the magnet coordinate system directions. However, when the MEG are correctly aligned with the beam directions the full displacement is measured in all cases allowing less pressure to be used when forming MR-ARFI images. This was confirmed in vivo by acquiring displacement maps with the MEG at different orientations (figure 3.5). When the MEGs are rotated 90° from the beam direction no focus can be detected. As the MEGs are brought into alignment with the beam direction a larger signal is detected.
Figure 3.6 shows a pressure vs displacement curve for MR-ARFI maps. Since one of the major sources of risk is cavitation, and this risk scales with PNP, finding a minimal pressure to reliably localize the focus is important. Using a free field PNP of 1.68 MPa, corresponding to an estimated in vivo PNP after skull transmission of 0.54 MPa, the focus could still be identified.
This pressure corresponds to a MI of 1.88 which is within the highest limits allowed by the FDA for ultrasound imaging [12], [46].
3.4—Discussion
Simulations and in vivo measurement were used to identified FUS parameters that can be used to transcranially induce displacements in brain tissue that can be measured by MR- ARFI. Using these parameters, displacement maps were measured in living macaque brains to
Figure 3.6: A: Displacement maps for increasing free field pressure sonications. Higher pressure sonications result in an easier to detect focus. B: The measured displacement increases with increasing PNP. Displacements were detectable with free field pressures that correspond to an MI below the highest acceptable FDA limits for diagnostic ultrasound.
localize the focus of the transducer safely and accurately. Much prior work has established MR- ARFI in phantoms [25]; this study demonstrates transcranial MR-ARFI in a survival imaging session in the brain of a large animal with intact skull with surrounding tissues of skin, soft tissue, and muscle without measurable negative bioeffects. This work shows the importance of parameter selection during the design of transcranial MR-ARFI protocols.
A low duty was used to avoid heating that could lead to adverse bioeffects in the brain, the skull, and the scalp. Heating was minimized by using the lowest FUS intensity needed to generate detectable displacement and separating the FUS pulses in time by lengthening MR repetition time to 1 second (overall duty cycle of 0.23%). The TR only needs to be short enough to acquire a large enough field of view to localize the focus. While tissue damage is known to occur with large temperature changes it is possible that even small temperature changes in the brain can temporally change neurological function, which could confound the results of a neuromodulation studies [120]. In both simulation and in vivo measurements we showed only small changes in temperature at the focus with our FUS pulses. In vivo MR thermometry is not able to directly measure temperature changes within the skull where simulations showed the largest temperature increase. By careful consideration of duty cycle and FUS intensity it is possible to localize the FUS beam with MR-ARFI with minimal heating.
In addition to the TR of the MR-ARFI sequence, other MR methods can be applied to reduce the needed risk of damage when localizing the focus. Improvements to the SNR of the sequence will allow for detection of smaller displacements that are generated with lower energy. Bipolar gradients have been shown to improve the phase images which are used to generate the displacement maps resulting in a more sensitive sequence [106]. Single shot EPI